Abundance of mass 47 CO2 in urban air, car exhaust, and human breath

Abundance of mass 47 CO2 in urban air, car exhaust, and human breath

Geochimica et Cosmochimica Acta 70 (2006) 1–12 www.elsevier.com/locate/gca Abundance of mass 47 CO2 in urban air, car exhaust, and human breath Hagit...

277KB Sizes 3 Downloads 13 Views

Geochimica et Cosmochimica Acta 70 (2006) 1–12 www.elsevier.com/locate/gca

Abundance of mass 47 CO2 in urban air, car exhaust, and human breath Hagit P. Affek *, John M. Eiler Division of Geological and Planetary Sciences, California Institute of Technology, MC100-23, Pasadena, CA 91125, USA Received 27 January 2005; accepted in revised form 16 August 2005

Abstract Atmospheric carbon dioxide is widely studied using records of CO2 mixing ratio, d13C and d18O. However, the number and variability of sources and sinks prevents these alone from uniquely defining the budget. Carbon dioxide having a mass of 47 u (principally 13 18 16 C O O) provides an additional constraint. In particular, the mass 47 anomaly (D47) can distinguish between CO2 produced by high temperature combustion processes vs. low temperature respiratory processes. D47 is defined as the abundance of mass 47 isotopologues in excess of that expected for a random distribution of isotopes, where random distribution means that the abundance of an isotopologue is the product of abundances of the isotopes it is composed of and is calculated based on the measured 13C and 18O values. In this study, we estimate the d13C (vs. VPDB), d18O (vs. VSMOW), d47, and D47 values of CO2 from car exhaust and from human breath, by constructing ÔKeeling plotsÕ using samples that are mixtures of ambient air and CO2 from these sources. d47 is defined as ðR47 =R47 std  1Þ  1000, where 47 13 18 R47 is the R value for a hypothetical CO whose d C = 0, d O = 0, and D = 0. Ambient air in Pasadena, CA, where this 2 VPDB VSMOW 47 std study was conducted, varied in [CO2] from 383 to 404 lmol mol1, in d13C and d18O from 9.2 to 10.2& and from 40.6 to 41.9&, respectively, in d47 from 32.5 to 33.9&, and in D47 from 0.73 to 0.96&. Air sampled at varying distances from a car exhaust pipe was enriched in a combustion source having a composition, as determined by a ÔKeeling plotÕ intercept, of 24.4 ± 0.2& for d13C (similar to the d13C of local gasoline), d18O of 29.9 ± 0.4&, d47 of 6.6 ± 0.6&, and D47 of 0.41 ± 0.03&. Both d18O and D47 values of the car exhaust end-member are consistent with that expected for thermodynamic equilibrium at200 C between CO2 and water generated by combustion of gasoline–air mixtures. Samples of CO2 from human breath were found to have d13C and d18O values broadly similar to those of car exhaust–air mixtures, 22.3 ± 0.2 and 34.3 ± 0.3&, respectively, and d47 of 13.4 ± 0.4&. D47 in human breath was 0.76 ± 0.03&, similar to that of ambient Pasadena air and higher than that of the car exhaust signature.  2005 Elsevier Inc. All rights reserved.

1. Introduction The budget of atmospheric CO2 is widely studied using records of mixing ratio, d13C and d18O, combined with estimates of d13C and d18O values associated with CO2 fluxes (Ciais et al., 1995a,b, 1997; Cuntz et al., 2003; Francey and Tans, 1987; Francey et al., 1995; Peylin et al., 1999). In particular, d13C values are primarily used to partition marine from terrestrial photosynthesis (Ciais et al., 1995a,b; Francey et al., 1995) and d18O values are used primarily to separate terrestrial net photosynthesis from soil

*

Corresponding author. Fax: +1 626 683 0621. E-mail addresses: [email protected] (H.P. Affek), [email protected]. edu (J.M. Eiler). 0016-7037/$ - see front matter  2005 Elsevier Inc. All rights reserved. doi:10.1016/j.gca.2005.08.021

respiration (Ciais et al., 1997; Peylin et al., 1999; Yakir and Wang, 1996). However, the number and variability of sources and sinks prevents these alone from uniquely defining the budget. For example, d13C values of respired CO2 and CO2 generated by combustion of oil, gasoline or biomass are effectively indistinguishable from one another. Previous efforts to separate their relative contributions to air using the d18O of CO2 have suffered from a paucity of data, and have resorted to the assumption that combustion sources are equal atmospheric O2 in d18O (Pataki et al., 2003a; Zimnoch et al., 2004). CO2 having a mass of 47 u (mainly 13C18O16O) provides an additional potential constraint on the atmospheric budget of carbon dioxide, but little is known in detail about its ability to differentiate among various budget components. Theoretical calculations and initial results suggest that

2

H.P. Affek, J.M. Eiler 70 (2006) 1–12

the mass 47 anomaly (D47) value of CO2, a measure of the excess (or deficit, if negative) of mass 47 isotopologues, relative to the abundance predicted for a random distribution of isotopes (a detailed definition is given below), is temperature dependent (Eiler and Schauble, 2004; Wang et al., 2004) and therefore might distinguish between high temperature processes, such as combustion, vs. low temperature processes, such as respiration. In this study, we report the isotopic composition, including mass 47 isotopologues, of CO2 in car exhaust and in air from an urban setting (Pasadena, CA), using a ÔKeeling plotÕ approach (Keeling, 1958, 1961). We then compare this combustion source to the isotopic composition of CO2 in human breath, taken as an example of low temperature respiratory processes.

In all cases, CO2 was extracted from collected air, exhaust or breath within several hours of sampling by slowly expanding the sample into a glass vacuum line through four liquid N2 traps. After all non-condensable gases were pumped away, liquid N2 traps were warmed in series by three ethanol-dry ice traps and finally the fourth trap was thawed to room temperature, liberating the collected CO2. The carbon dioxide was then transferred into a calibrated volume, where its pressure was measured using an MKS Baratron, allowing us to calculate the number of moles of collected CO2. This number of moles was then re-calculated as CO2 mixing ratio based on the known volumes of the sample flasks and measurements of pressure in the sample flasks prior to CO2 extraction. 2.2. Isotopic analysis

2. Methods 2.1. Air sampling and CO2 extraction Air was sampled at a flow rate of 1 L min1 through a Mg(ClO4)2 trap into pre-evacuated 5-L glass flasks. The Mg(ClO4)2 trap reduced the moisture content in car exhaust from approximately 80% relative humidity to 1–2% (measured using a Li-Cor 1400-104 relative humidity sensor). ‘‘Ambient’’ air was sampled in the campus of the California Institute of Science (Caltech) in Pasadena, CA, between April and July 2004, always at 10 am. Samples were obtained from a third floor balcony of the North Mudd building on the Caltech campus (34.1365N, 118.1275W). These samples invariably contain a small, variable fraction (ca. 15–30 lmol mol1) of CO2 from local fossil-fuel sources (Newman et al., 1999). ‘‘Traffic’’ air was sampled at a time of relatively heavy traffic (8 am) at street level at the intersection of E. California and S. Wilson in Pasadena, CA (34.1360N, 118.1279W). ‘‘Exhaust’’ air was sampled at distances of 5–200 cm from the exhaust pipe of either of two cars (most samples are exhaust from a 2001 Ford Taurus; the others are from a 1991 Jeep Cherokee; Table 1). Samples of ‘‘ambient,’’ ‘‘traffic,’’ and ‘‘exhaust’’ air were used in a ÔKeeling plotÕ analysis to characterize the isotopic composition of the car exhaust end-member. Samples were collected over the course of 1 h, in 3 different days. Variations in the background air CO2 mixing ratio and isotopic composition are likely to occur among sampling days but since the ÔKeeling plotsÕ include samples of high CO2 mixing ratios, close to those of the end-member, the effect of variations in the background on the intercept is insignificant. Several additional exhaust samples were obtained a year later in a similar fashion through a 1/4 in. stainless steel tube inserted 15– 40 cm into the tail pipe of the Ford Taurus (denoted ‘‘tail pipe CO2,’’ Table 2). Human breath was sampled from four persons by exhaling through a Mg(ClO4)2 trap into pre-evacuated 500 ml flasks. For one of these persons, mixtures of ambient air and respired CO2 ware obtained by exhaling at various distances from the flask inlet.

The isotopic composition of extracted CO2 was measured within 3 days of extraction by dual inlet gas source mass spectrometry using a Finnigan-MAT 253 isotope ratio mass spectrometer, configured to measure masses 44, 45, 46, 47, 48, and 49. A detailed description of the mass spectrometer is given by Eiler and Schauble (2004). Values of d13C and d18O were standardized by comparison with CO2 generated by phosphoric acid digestion of NBS-19 and are reported vs. VPDB and VSMOW, respectively. In reporting the abundance of mass 47 CO2 we use the variable D47, referred to here as the mass 47 anomaly. The anomaly for any isotopologue mass, i, Di, is defined as the difference (in per mil) between the measured Ri (=[mass i]/[mass 44], where i refers to masses 45, 46, or 47) and that ratio expected for CO2 whose isotopes are distributed randomly among all isotopologues (Eiler and Schauble, 2004; Wang et al., 2004). Random distribution means that the abundance of an isotopologue is the product of the abundances of the isotopes it is composed of. For example, [13C18O16O]* = 2[13C][18O][16O], where * denotes the random distribution, and 2 is a symmetry number indicating that 18O can sit in either of two equivalent positions. The randomly distributed abundance of mass 47 is then [47]* = [13C18O16O]* + [12C18O17O]* + [13C17O17O]* and D47 is defined as  D47 ¼

  46   45  R47 R R  1   1   1  1000; R47 R46 R45 ð1Þ

where R47, R46, and R45 are measured values. Ri* for every sample is calculated from the measured R13 and R18 (see Eiler and Schauble, 2004, and Appendix B for more details and for an example of the calculation). R17 is calculated from R18 0:5164 assuming the relationship R17 ¼ ðR18 =R18  VSMOW Þ 17 18 17 RVSMOW , where RVSMOW and RVSMOW are taken as 0.0020052 and 0.0003799, respectively (Gonfiantini et al., 1993). Note that D47 values are only slightly sensitive to the power used to calculate R17, with variation of ±0.002& in D47 for power between 0.5 and 0.528 (i.e., instead of 0.5164). Measure-

Mass 47 CO2 in urban air, car exhaust, and human breath

3

Table 1 CO2 mixing ratios and isotopic composition in air, car exhaust, and breath samples (average ± 1r, n = 3–6, no error for samples measured once) CO2 (lmol mol1)

d13CVPDB (&)

d18OVSMOW (&)

d47 (&)

D47 (&)

Exhaust air 22274 19898 3468 3371 3312a 1528 1439 776 574 437a

24.205 ± 0.010 24.105 ± 0.008 22.803 ± 0.016 22.885 ± 0.004 22.898 20.276 ± 0.053 20.371 ± 0.024 16.685 ± 0.003 13.413 ± 0.012 10.135

29.974 ± 0.025 30.128 ± 0.013 30.919 ± 0.026 30.914 ± 0.005 29.093 33.731 ± 0.116 34.803 ± 0.025 36.092 ± 0.007 38.772 ± 0.019 40.959

6.868 ± 0.046 7.060 ± 0.015 9.250 ± 0.072 9.102 ± 0.081 7.236 14.715 ± 0.209 15.712 ± 0.063 20.699 ± 0.027 26.704 ± 0.060 32.377

0.439 ± 0.051 0.376 ± 0.032 0.473 ± 0.044 0.423 ± 0.073 0.383 0.621 ± 0.042 0.596 ± 0.047 0.658 ± 0.030 0.707 ± 0.032 0.866

Traffic air 446 424 420 416 412

11.964 10.933 12.311 10.313 10.383

39.524 40.542 39.151 41.188 41.260

28.968 31.073 28.292 32.417 32.448

0.736 0.777 0.760 0.848 0.868

Ambient air 404 404 401 396 384 383

10.192 ± 0.033 10.001 ± 0.009 9.199 ± 0.011 9.653 ± 0.002 9.340 ± 0.004 9.769 ± 0.009

41.883 ± 0.076 41.719 ± 0.010 41.579 ± 0.040 41.078 ± 0.033 40.599 ± 0.018 40.641 ± 0.018

33.346 ± 0.075 33.128 ± 0.027 33.914 ± 0.073 32.977 ± 0.064 32.809 ± 0.063 32.461 ± 0.035

0.956 ± 0.048 0.732 ± 0.040 0.841 ± 0.042 0.867 ± 0.032 0.871 ± 0.055 0.909 ± 0.007

Human breath 30333b 29287b 19889c 7323b 5184d 5055e 1142b

23.086 ± 0.006 21.875 ± 0.002 23.121 ± 0.003 21.015 ± 0.008 21.201 ± 0.103 20.678 ± 0.020 17.358 ± 0.006

34.299 ± 0.016 34.549 ± 0.002 34.483 ± 0.008 35.379 ± 0.017 35.065 ± 0.216 33.225 ± 0.080 36.959 ± 0.009

12.632 ± 0.024 14.153 ± 0.016 12.793 ± 0.042 15.971 ± 0.024 15.212 ± 0.282 13.861 ± 0.150 20.952 ± 0.075

0.772 ± 0.006 0.768 ± 0.015 0.783 ± 0.033 0.930 ± 0.033 0.710 ± 0.038 0.681 ± 0.114 0.749 ± 0.002

All data are corrected for N2O abundance. See Appendix B for an example of D47 and d47 calculation procedure. Samples were collected between April and July 2004. a Exhaust air samples from Jeep Cherokee. The rest of exhaust samples are from Ford Taurus. b Breath of person 1. The samples of 7323 and 1142 lmol mol1 are mixtures of breath and ambient air instead of direct breath samples. c,d,e Breath from persons 2, 3, and 4 respectively.

Table 2 Isotopic composition of ‘‘tail pipe CO2’’ collected at depths of 15–40 cm in the tail pipe of a Ford Taurus 2001 (average ± 1r, n = 5–6) Depth in tail pipe (cm)

d13CVPDB(&)

d18OVSMOW (&)

D47 (&)

15 20 25 30 40

25.168 ± 0.002 24.363 ± 0.004 24.858 ± 0.005 24.911 ± 0.011 25.026 ± 0.008

29.737 ± 0.007 30.624 ± 0.012 29.993 ± 0.011 29.893 ± 0.018 30.374 ± 0.019

0.460 ± 0.044 0.346 ± 0.029 0.380 ± 0.047 0.341 ± 0.057 0.361 ± 0.031

All data are corrected for N2O abundance. See Appendix B for an example of D47 calculation procedure. Samples were collected in June 2005.

ments of D47 are standardized by comparison with CO2 that had been heated to 1000 C for 2 h, driving it to the random distribution (D47 ” 0). Conservative mixing does not necessarily lead to linear trends in plots of D47 vs. 1/[CO2] or vs. d13C or d18O, for reasons discussed in Eiler and Schauble (2004), Zyakun

(2003), Kaiser et al. (2003), and in more detail in Appendix A of this work. Therefore, we find it useful to also report the d47 value, which is defined here as 47 47  1Þ  1000, where R is the R value for a ðR47 =R47 std std hypothetical CO2 whose d13CVPDB = 0, d18OVSMOW = 0, and D47 = 0. Since d47 is defined using the traditional delta notation, it produces linear trends in a ÔKeeling plotÕ for conservative mixing processes (i.e., like d13C and d18O). The measured R47 value can be influenced by interferences from hydrocarbons and/or their fragmentation/recombination products (Eiler and Schauble, 2004). Therefore, prior to analysis the CO2 was further purified by passing it through a Chrompack Poraplot-Q GC column, 25 m long, 0.32 lm ID, held at room temperature with 3 ml min1 helium as carrier. This purification method was later improved by using a Supelco QPlot column, 30 m long, 0.53 mm ID, held at 10 C with 2 ml min1 helium as carrier and baked at 150 C between samples. CO2 was collected in a glass trap immersed in liquid N2 at the end

4

H.P. Affek, J.M. Eiler 70 (2006) 1–12

of the column. The extent of remaining hydrocarbon contamination was estimated for each sample by monitoring the signal for mass 49, which is highly correlated with mass 47 excesses resulting from many contaminants (Eiler and Schauble, 2004). A sample was considered satisfactorily clean if there was no correlation between D47 values observed in five measurements and the difference in the mass 49 signal between sample and reference gas. This was typically obtained when the sample mass 49 signal was up to 0.5 mV (read through a 1012 X resistor) higher than the reference gas signal. For car exhaust samples a cycle of GC cleaning as described above was typically not enough to bring the mass 49 signal of the sample to a satisfactorily low value and the CO2 was typically passed a second time through the GC column (this was not necessary in the later improved GC method). Mass spectrometry precision was typically 0.01& for d13C, 0.03& for d18O, 0.07& for d47, and 0.04& for D47 (all 1r) for a typical sample size of 50–100 lmol CO2 that generated 10 V signal for mass 44 (read through a 3 · 108 X resistor) at typical instrumental conditions. This precision estimate was based on replicate mass spectrometry measurements of clean CO2. External precision for replicate air samples is poorer due to variable contamination and fractionation during sample preparation. We estimated this precision by replicate extractions of CO2 from a cylinder of compressed air followed by GC removal of hydrocarbons and mass spectrometry. D47 was calculated for each mass spectrometer run, averaged for each sample (the standard deviation of this measurement gave the mass spectrometry precision), and compared among CO2 samples to give external precision. External precision averaged 0.1& for d13C, 0.3& for d18O, 0.4& for d47, and 0.07& for D47. The precision in D47 measurements is better than that in d47 since the errors due to fractionation in d13C, d18O, and d47 are correlated. All measurements of d13C, d18O, d47, and D47 have been corrected for interferences from isotopologues of N2O, which is not removed by our sample preparation method. The amount of N2O in each sample was estimated by measuring the D14 value, defined as: D14 = (V14s  V14r)/ ((V44s + V44r)/2), where V14 and V44 are the mass 14 and mass 44 voltages, respectively, and subscripts s and r refer to the sample and reference, respectively. The relationships between this variable and the isotopic values were calibrated by analyzing mixtures of N2O and CO2 prepared by freezing measured amounts of the pure gases together in liquid N2. N2O corrections for air samples averaged 0.19 ± 0.03& for d13C, 0.27 ± 0.04& for d18O (average ± 1r, n = 7), comparable to commonly used corrections (Mook and Jongsma, 1987; Mook and Van Der Hoek, 1983), 0.34 ± 0.05& for d47 and +0.086 ± 0.014& for D47. We also tested the efficacy of N2O corrections based on measurements of mass 30 (i.e., 14N16O). We compared N2O correction using mass 14 to the correction using mass 30 by measuring air samples as well as the same

samples spiked with small amounts of N2O. When corrected using mass 14 the isotopic values were similar in the original air sample and in the sample spiked with N2O, whereas the correction based on mass 30 gave a significant deviation (probably because of a high mass 30 signal for CO2, possibly due to the presence of 12C18O, as opposed to very low mass 14 signal). A potential problem in the use of mass 14 is interference due to trace amounts of N2 in the sample. This was avoided by numerous steps (typically more than 10·) of freezing the CO2 and N2O in liquid nitrogen and pumping away any non-condensable gases. Furthermore, the integrity of the measurement was verified by observing D14 = 0 when measuring a standard containing no N2O. Taking into account variations between 40 and +40& in d15N and between 20 and 60& in d18O values of N2O (Rahn, 2005) would lead to a range of 0.02& in the possible N2O correction for D47. This is insignificant considering the precision of our measurement. Gasoline was sampled at the Union 76 gasoline station, right after fueling the car used for most of the exhaust measurements (2001 Ford Taurus) by collecting several ml of gasoline from the same pump and immediately transferring it into closed glass vials. d13C in gasoline was measured by combustion in a Carlo Erba elemental analyzer (NA1500NC) attached to a Delta Plus Finnigan mass spectrometer at UC-Irvine. It is difficult to prevent evaporation of the gasoline samples while preparing the sample for measurement, and this might lead to fractionation of d13C values. To test the potential effect of evaporation, we measured samples of various sizes (1.3–2.3 mg) and samples were kept a variable amount of time (5–15 min) before measurement, allowing variable degrees of evaporation from the closed capsules (of up to 50% evaporated). No significant differences in d13C were observed suggesting no significant fractionation in evaporation, consistent with data from Wang and Huang (2003) for C10 to C14 nalkanes. 2.3. Statistical analysis Estimated errors in sample handling and mass spectrometry analysis are given as the standard deviation about the mean (1r). Statistical analyses of trends in ÔKeeling plotsÕ presented in Section 3 were made using the regression function in the data analysis add-in of Microsoft Excel X for Macintosh, where the error estimate for the concentrated end-member defined by the ÔKeeling plotÕ trend is the standard error of the intercept. This analysis is a standard linear regression calculation that assumes no error associated with the x variable (model I, Miller and Tans, 2003; Pataki et al., 2003b). Regression calculation for ÔKeeling plotÕ end-members is also performed using a second method, including uncertainty in the x variable (model II, assuming 1% error in the concentration values), according to the method of Reed (1992) by iterations using the Solver function in Microsoft Excel X for Macintosh.

Mass 47 CO2 in urban air, car exhaust, and human breath

3. Results and discussion 3.1. Car exhaust and urban air Ambient air sampled on the Caltech campus varied in CO2 mixing ratios from 383 to 404 lmol mol1, in d13C from 9.2 to 10.2&, in d18O from 40.6 to 41.9&, in d47 from 32.5 to 33.9&, and in mass 47 anomaly (D47) from 0.73 to 0.96& (Table 1 and see Appendix B for a detailed example of calculating D47 and d47). The CO2 mixing ratios and bulk isotopic compositions (d13C and d18O values) observed in Pasadena ‘‘ambient’’ air were within the range of those previously observed in a several year record (Newman et al., 1999) and in spring 2003 measurements in the Los Angeles basin (Eiler and Schauble, 2004). The mixing ratios are higher and d13C values are lower than those of global background air (i.e., far from point sources or sinks). They are similar to those of suburban air in Paris (Widory and Javoy, 2003) and urban air in Dallas, Texas (Clark-Thorne and Yapp, 2003) and Krakow, Poland (Zimnoch et al., 2004), suggesting anthropogenic contributions even to the lowest mixing ratio samples. Values of d18O observed here are slightly higher than those observed in Salt Lake City, Utah (Pataki et al., 2003a). Values of D47 observed here are slightly higher, on average, than those observed in spring of 2003 in Pasadena (0.62–0.75& with one measurement of 0.93&; Eiler and Schauble, 2004). ‘‘Traffic’’ air had consistently higher CO2 mixing ratios, from 412 to 446 lmol mol1, similar to values observed in the streets of Paris (Widory and Javoy, 2003). It varied in d13C from 10.3 to 12.3&, in d18O from 39.2 to

5

41.2&, in d47 from 28.3 to 32.5&, and in D47 from 0.74 to 0.87&. ‘‘Exhaust’’ air samples collected 30 or more cm from the car exhaust pipe varied in CO2 mixing ratios from 574 to 3468 lmol mol1, in d13C from 13.4 to 22.8&, in d18O from 38.8 to 30.9&, in d47 from 26.7 to 9.3&, and in D47 from 0.71 to 0.47&. Two samples obtained 5 cm below the car exhaust tail pipe had CO2 mixing ratios of 2 and 2.2%, d13C of 24.2 and 24.1&, d18O of 30.0 and 30.1&, d47 of 6.9 and 7.1&, and D47 of 0.44 and 0.38&. These three groups of samples (‘‘Ambient,’’ ‘‘Traffic,’’ and ‘‘Exhaust’’) were used together to derive the isotopic composition of the CO2 emitted from cars using a ÔKeeling plotÕ mass balance approach (Fig. 1). When analyzing the plot by standard linear regression (model I), the intercept for the trend defined by all measurements, reflecting the car exhaust end-member, is characterized by a d13C value of 24.4 ± 0.2&, d18O of 29.9 ± 0.4&, d47 of 6.6 ± 0.6&, and D47 of 0.41 ± 0.03& (intercept ± SE). No significant change was observed in the end-member values when the regression calculation included uncertainty in the x variable (see Appendix A). We therefore use the above values in the following discussion. D47 is not linear in simple mixing but for mixing of car exhaust into air this non-linearity has only a small effect (smaller than the precision analysis) which we ignore in the following discussion (see Appendix A for details). The d13C of CO2 in car exhaust primarily depends on the fuel used. Petroleum used to produce gasoline varies in d13C between 29.9 and 31.5& for non-marine sources, and between 23.1 and 32.5& for marine sources (Yeh and Epstein, 1981), with a global weighted average

Fig. 1. ÔKeeling plotsÕ for d13C (vs. VPDB), d18O (vs. VSMOW), d47 and D47 in CO2. [CO2] is the CO2 mixing ratio in lmol mol1. Empty symbols refer to CO2 in air mixed with car exhaust obtained by sampling at varying distance from car exhaust pipe as well as air at a busy junction and ‘‘ambient’’ air in the Caltech campus in Pasadena, CA. Bold symbols refer to human breath samples and air mixed with human breath. The obtained ÔKeeling plotÕ trends were d13C = 5775/[CO2]  24.4, d18O = 4533/[CO2] + 29.9, d47 = 10498/[CO2] + 6.6, and D47 = 176/[CO2] + 0.41 for car exhaust, and d13C = 5002/ [CO2]  22.3, d18O = 2759/[CO2] + 34.3, d47 = 7810/[CO2] + 13.4, and D47 = 38/[CO2] + 0.76 for human breath. See Appendix A for a discussion on linearity approximation for the D47 mixing line.

6

H.P. Affek, J.M. Eiler 70 (2006) 1–12

estimated to be 26.5& (Andres et al., 2000; Tans, 1981). The d13C of the gasoline used to fill the tank of the car used for most of our ‘‘Exhaust’’ data points was measured to be 24.4 ± 0.2& (n = 4), in agreement with the car exhaust end-member defined by our data. This value is higher than those typically observed for gasoline (Clark-Thorne and Yapp, 2003; Pataki et al., 2003a; Widory and Javoy, 2003) but is similar to that observed in Poland (25&, Zimnoch et al., 2004) and can be explained by the mixture of sources of petroleum products used in California (approximately half local sources, and the rest Alaskan and imported oil; Sally Newman, personal communication, 2005). Most urban air measurements result in an anthropogenic end-member that is lower in d13C than gasoline due to the contribution of CO2 produced from combustion of natural gas (Pataki et al., 2003a; Takahashi et al., 2002; Zondervan and Meijer, 1996 and Sally Newman, personal communication, 2005). However, when measured in streets with significant automobile traffic, as in the case of our ‘‘Traffic’’ and ‘‘Exhaust’’ measurements, the anthropogenic end-member closely approaches the d13C of local gasoline (Widory and Javoy, 2003). In global models of the isotopic budget of CO2, its d18O value is dominated by exchange with plant and soil water, and the anthropogenic component is generally neglected (Ishizawa et al., 2002). In urban air and in some continental stations, however, the anthropogenic component is likely to be significant. The d18O of CO2 produced in combustion is generally assumed to equal that of atmospheric oxygen (Ciais et al., 1997; Pataki et al., 2003a; Peylin et al., 1999; Zimnoch et al., 2004; Zondervan and Meijer, 1996); that is, 23.8& (Coplen et al., 2002; although 23.5& (Kroopnick and Craig, 1972) is used in most studies of urban CO2 sources). In Salt Lake city, the weighted average d18O values of urban CO2 sources, at times when combustion of either gasoline or natural gas is expected to be the dominant source, were found to be slightly lower than this value (21.9& for rush hour measurements and 23.1& for winter nights). The difference was attributed to a contribution from soil respiration (Pataki et al., 2003a). In Pasadena, on the other hand, the weighted average d18O of urban CO2 sources has been estimated at 27& (Newman et al., 1999). This value is closer to our car exhaust end-member (29.9 ± 0.4&), but still significantly lower. In this case, the difference might reflect contributions from natural gas combustion, producing CO2 with a d18O equal to that of O2 in air, or from soil respiration, which we estimate should have a d18O of approximately 24& (local tap water is 9&, Lisa Welp, personal communication, 2004; and soil respiration flux should be approximately 7& lower than equilibrium with this water, Miller et al., 1999). We expected CO2 produced in car exhaust to have a D47 value close to zero as was observed in one of two car exhaust samples measured by Eiler and Schauble (2004), because it is generated by combustion at temperatures of at least 900 C. Instead, the measured D47 value of 0.41 ± 0.03& is consistent with the equilibrium distribu-

tion of isotopologues at 195 ± 15 C (Wang et al., 2004). We suggest this reflects isotopic re-equilibration of CO2 in the car exhaust stream, mediated by exchange with water, as detailed in the following paragraph. Isotopic exchange among isotopologues of pure gaseous CO2 is slow relative to the timescales of exhaust production and emission, even at temperatures of up to several hundred degree celsius. However, in the presence of condensed water (at near ambient temperatures) or water vapor (at elevated temperatures), re-equilibration of CO2 isotopologues is far faster, presumably facilitated by CO2–H2O exchange. In this case, CO2 re-equilibration must be accompanied by changes in its d18O value, by an amount that depends on the relative amounts of CO2 and H2O, and the temperature of exchange. The temperature dependency of oxygen isotopic fractionation between CO2 and liquid H2O is summarized in Friedman and OÕNeil (1977, Data from Figs. 5 and 6 of that reference up to 300 C were used to estimate fractionation at higher temperatures by fitting a second order polynomial of 1/T (K) and extrapolating to high temperature by forcing the fractionation factor to zero at 1/T = 0, giving a correlation of R2 = 0.999). The fractionation between CO2 and water vapor (eCO2 –H2 O in Eq. (2)) has not been previously measured at temperatures relevant to our calculation, but can be estimated as the sum of the fractionation factors in exchange between CO2 and liquid water, and in evaporation of water (values for water evaporation are taken from curve A in Fig. 8 in Friedman and OÕ Neil, 1977 and are forced to zero above the critical point of 374 C). The resulting gas-phase fractionation factor at 195 ± 15 C (the temperature of reequilibration implied by the D47 value of exhaust CO2) is 20.5 ± 1.3&. If we assume that gasoline combustion is the only source of water in the exhaust gas (and that water and CO2 are the only oxygen bearing combustion products present in significant amounts), we can use the stoichiometry of the components of gasoline to estimate the relative amount of CO2 and H2O in the exhaust gas and therefore of the distribution of oxygen used in combustion between CO2 and H2O. The C/H ratio of gasoline (0.56, estimated from the weighted average stoichiometry of the components of gasoline, Na et al., 2004) implies that the fraction of oxygen atoms, which is present in CO2 ðXO–CO2 ¼ 2C=ð2C þ 0:5HÞÞ, is 0.69. We further assume that all oxygen in both CO2 and H2O comes from O2 in air, and thus that the weighted average d18O of the exhaust stream is 23.8&. Given these constraints and assumptions, we can calculate the oxygen isotope mass balance for CO2 and H2O in the exhaust stream using the relationship d18 OO2 ¼ XO–CO2 d18 OCO2 þ XO–H2 O d18 OH2 O ¼ XO–CO2 d18 OCO2

 18  d OCO2  eCO2 H2 O e  . þ ð1  XO–CO2 Þ CO2 –H2 O þ1 1000

ð2Þ

Mass 47 CO2 in urban air, car exhaust, and human breath

Substituting the constraints quoted above into this relationship yields a predicted d18O for CO2 in exhaust of between 30.6& (for re-equilibration at 210 C) and 31.6& (for re-equilibration at 180 C), which closely approaches the measured d18O value of the car exhaust end-member (Fig. 2). We conclude that both the D47 and d18O values of CO2 in car exhaust can be understood as a consequence of isotopic re-equilibration between CO2 and H2O in the exhaust stream, quenching at ca. 200 C. The exhaust does not cool to such a temperature until it exits the catalytic converter, and so presumably this quenching occurs between the downstream side of the catalytic converter and the tail pipe. Alternatively, both D47 and d18O values in the car exhaust may also be explained by only partial exchange with exhaust water at a lower temperature. For example, a mixture of approximately equal proportions of CO2 equilibrated at 900 C with CO2 re-equilibrated at 50 C (the temperature of the exhaust at the exit of the tail pipe) would yield D47 and d18O values consistent with a temperature of 200 C. In an attempt to resolve these two alternative interpretations, we sampled exhaust gas at depths of 15–40 cm within the tail pipe (Table 2. These measurements were performed a year after the above data were obtained, when gasoline composition was potentially different, and are therefore not included in the ÔKeeling plotsÕ). We anticipate that if the measured D47 reflects full exchange at 200 C, we should find a constant D47 and d18O in all parts of the tail pipe having temperatures lower than 200 C. In contrast, if the measured D47 reflects mixing of high and low temperature gas, we should find a gradient of increasing D47 and d18O with decreasing temperature of collection, or values near D47 = 0& at all temperatures above the partial exchange temperature. We found that at depths up to

7

40 cm, corresponding to collection temperatures up to varied between 0.34 ± 0.06& and 70 C, D47 0.46 ± 0.04&, and d18O varied from 29.7 ± 0.01& to 30.6 ± 0.01& with no clear spatial gradient (Table 2). Although inconclusive, these results support the interpretation that CO2 is homogeneously reset to 200 C. Note that exchange in sample flasks after sampling would result in a D47 value of approximately 0.95&, irrespective of the amount of water available, inconsistent with our results. As opposed to the two cars measured in this work, one sample measured previously (Eiler and Schauble, 2004) had lower values for both D47 and d18O (0.06 ± 0.08& and 24.24 ± 0.18&, respectively). These lower values are consistent with Eq. (2) for the expected combustion temperatures (D47 = 0.06 ± 0.08& corresponds to approximately 700 C with the error bars between 450 C and more than 1000 C). It is conceivable that cars of different engine designs and possibly even different driving conditions would have different re-equilibration temperatures and therefore different values for both D47 and d18O, though we expect that these two isotopic tracers should stay correlated through Eq. (2). In other words, when incorporating D47 values into isoflux models one should not use a single value for D47 and for d18O of combustion, but an equation that ties both tracers together (Fig. 2). Furthermore, it seems likely that other combustion processes, such as natural gas combustion in a power plant where the path of the flue gas in the chimney is different than that of car exhaust, could lead to different re-equilibration temperatures and different combined D47–d18O signatures following Eq. (2). We use Eq. (2) to calculate the d18O of the CO2 in equilibrium with water produced in combustion of natural gas (XO–CO2 ¼ 0:50 based on the stoichiometry of methane). The higher hydrogen content of natural gas yields additional water in combustion, thus leading to a higher d18O value for the CO2 at any given exchange temperature (Fig. 2). Therefore, D47 and d18O of CO2 produced in combustion should be included together in models not as constant values but through equations describing the relationships between these two tracers, where different kinds of fuel (i.e., gasoline and natural gas) would each have a specific equation. 3.2. Human breath

Fig. 2. Correlation between D47 and d18O (vs. VSMOW) of CO2 produced in combustion and undergoing isotopic exchange with water produced in combustion at different temperatures. d18O was estimated by a mass balance of the oxygen atoms (Eq. (2)) using the isotopic fractionation factor in 18O exchange between CO2 and water vapor, estimated as the sum of the fractionation factors in water evaporation and CO2 exchange with liquid water (raw data from Friedman and OÕ Neil, 1977) and a fraction of oxygen atoms in CO2 estimated from the stoichiometry of chemical components of gasoline (Na et al., 2004) and methane (as representing natural gas). Full line depicts gasoline and dashed line depicts methane. The gray rectangle depicts the range of car exhaust end-members from Fig. 1.

The CO2 mixing ratios in breath samples of four individuals varied between 0.5 and 3% (Table 1). These breath samples varied in d13C from 20.7 to 23.1&, in d18O from 33.2 to 35.1&, in D47 from 0.68 to 0.78&, and in d47 from 12.7 to 15.2&. Mixing lines between human breath and background air on a ÔKeeling plotÕ were used to estimate respiratory end-members. Mixing lines for d13C, d18O, and d47 were largely similar to those obtained for car exhaust mixing into air (Fig. 1), with standard linear regression calculations (model I) yielding end-member values of 22.3 ± 0.2&, 34.3 ± 0.3&, and 13.4 ± 0.4&, respectively. D47 in breath samples, on the other hand,

8

H.P. Affek, J.M. Eiler 70 (2006) 1–12

was not significantly different than that of background air, averaging 0.74 ± 0.05& (average ± 1r). CO2 in human and animal breath is typically slightly lower in d13C than body organic matter, which in turn is slightly higher in d13C than dietary carbon (De Niro and Epstein, 1978). Therefore, variations in breath d13C among individuals and regions are likely to be due to variations in diet. Values of d13C observed here are within the range previously observed in this region (Epstein and Zeiri, 1988) and are similar to those observed recently in the area (Eiler and Schauble, 2004) and to an experiment of controlled diet based on C3 plants but supplemented with C4 glucose (Panteleev et al., 1999). The values we observe are slightly higher than the values observed in Paris, France (24.5&, Widory and Javoy, 2003), suggesting that the diet in Southern California is slightly higher in d13C than that in France, possibly due to a larger contribution of C4 plants to the Californian diet. The d18O value of CO2 in animal breath is in equilibrium with body water, and is approximately 38& higher than body water (Luz et al., 1984), which in turn is related to drinking water (Luz et al., 1984). The d18O values observed here for human breath are within the range observed in the past in the region (Eiler and Schauble, 2004; Epstein and Zeiri, 1988) and close to the value expected based on local tap water d18O of approximately 9&. Eiler and Schauble (2004) observed a preliminary value for D47 in human breath of 0.66&, slightly lower than the value observed in the current study (0.74&). The current value of 0.74& is closer to, though still slightly lower than the value predicted based on equilibrium with water at body temperature (0.89& predicted for 37 C, Wang et al., 2004), suggesting that D47 values of human breath, and possibly other respiration processes, are affected by an additional unknown process and do not carry a pure equilibrium signal. The key for our purposes, however, is that D47 in breath is within measurement error of that in ambient air samples and differs significantly from that of car exhaust. The d13C of respiration (as exemplified by human breath) is similar to that of CO2 produced by combustion and is therefore not very useful in distinguishing between the two CO2 sources. Values of d18O could be used to separate respiration from combustion sources, although the respiration source is expected to vary substantially with the d18O of local meteoric water, and in many regions overlaps the d18O we observe here for car exhaust, again making it difficult to de-convolve the two sources using d18O. Values of d47, the variable newly defined here, should be highly correlated with d18O in combustion and respiration sources. In contrast, the mass 47 anomaly (D47) for CO2 that undergoes exchange with water at ambient temperatures, as in human, plant, and soil respiration, is predicted to vary only due to equilibrium temperature (Wang et al., 2004), and to be uniformly high (ca. 0.8&) compared to combustion sources (0.4& for exhaust in this study; 0.0& for natural gas combustion in Eiler and Schauble (2004)). Thus, the isotopic index D47 can be used to distin-

guish CO2 from combustion sources vs. respiration sources (and other sources involving low-temperature isotopic equilibration with water). Acknowledgments We thank Xiaomei Xu of UC-Irvine for measurements of the d13C of gasoline, and Sally Newman, Lisa Welp, and Zhengrong Wang for helpful discussions as well as James Farquhar, Jan Kaiser, and two anonymous reviewers for constructive comments and suggestions. This work is supported by Grants NSF-EAR-0091842, NSF-EAR0220066, and NSF-EAR-0345905. Associate editor: James Farquhar

Appendix A. Non-linearity of trends in ÔKeeling plotsÕ based on D47 The ÔKeeling plotÕ approach for finding mixture endmembers in air relies on the linearity of the delta scale to mixing. However, conservative mixing leads to non-linear variations in mass 47 anomaly (Eiler and Schauble, 2004). In this appendix, we examine this problem in detail for mixtures between CO2 end-members having bulk stable isotope compositions similar to those of natural sources. First, samples of CO2 having a range of initial bulk stable isotope compositions were heated to 1000 C, to drive their D47 values to 0.0&. We describe these samples as Ôheated gases.Õ We then prepared a second set of samples of CO2 having a range of initial bulk stable isotope compositions by exposing them to liquid water at room temperature for several days to drive their D47 value toward the roomtemperature equilibrium of ca. 0.9&. We describe these samples as Ônon-heated gases.Õ Fig. 3A shows the isotopic compositions of various mixtures of a heated and a non-heated gas that were similar to one another in bulk stable isotope composition (d13C values of 34.0 and 33.9&; d18O values of 4.0 and 4.9&; D47 values of 0.907 and 0&). Fig. 3B shows the isotopic compositions of various mixtures of a heated and a nonheated gas that were strongly different from one another in bulk stable isotope composition (d13C values of 3.6 and 33.4&; d18O values of 39.7 and 5.7&; D47 values of 0.828 and 0&). Data trends in Figs. 3A and B are compared to two types of calculated mixing lines: one that assumes conservative linear mixing of D47 values (incorrect, but a useful visual reference), and a second that assumes linear mixing of the measured delta values (d13C, d18O as well as raw d45, d46, and d47 values) and then calculates the D47 value of each mixture based on these mass balance predicted delta values. When the two end-members are similar in bulk isotopic composition, no significant difference is observed between the two mixing lines; that is, D47 values can be treated as if they were a conservative tracer in mixing problems (Fig. 3A, note that the two mixing lines fall on top of

Mass 47 CO2 in urban air, car exhaust, and human breath

9

exhaust sample of highest mixing ratio was arithmetically mixed with an average air sample to obtain the mass balance predicted d13C, d18O as well as d45, d46, d47 values and the fully calculated D47 values (as above), and compared to the mass balance predicted D47 (assuming linearity in conservative mixing of D47 values). The difference between fully calculated D47 and that calculated assuming linearity reached a maximum of 0.042&, for the 1:1 mixture (Fig. 3C). As with the artificial mixture (Fig. 3B), the fully calculated D47 in Fig. 3C can also be fitted with a second order polynomial function, whose intercept reflects the correct D47 of car exhaust end-member. The difference between the endmember obtained using the polynomial fit and that obtained using a linear fit to the fully calculated D47 was 0.025& (D47 = 0.39& for the polynomial fit). This deviation from linearity is too small to be observed in Fig. 1, since it is at the limit of the measurement accuracy. However, applying a mass balance approach to future air sample analysis, as is done for d13C and d18O, would likely require improved accuracy than is currently available and would then also require considering the non-linearity of D47. The non-linearity is explained by the mathematical behavior of D47 in a mass balance treatment. D47 is defined and calculated as  47   46   45  R R R D47 ¼ 1  1   1  1000; R47 R46 R45 ðA1Þ

Fig. 3. Nonlinearity of D47 in mixing. Curves are calculated mixing lines using heated (D47 = 0) and non-heated end-members (D47 = 0.907 in (A) and 0.828 in (B)). Long dashed line is the mixing line assuming conservative mixing of D47 values. Full line is a polynomial fit to the fully calculated D47 (assuming linearity in measured d values but not in mass 47 anomaly). Short dashed line is a linear fit to fully calculated D47. Bold circles are measured D47 values. (A) Mixing lines for end-members similar in d13C and d18O (note that all lines seem as one as they are practically identical for this mixture). (B) Mixing lines for end-members of different d13C and d18O. (C) Arithmetic mixing of car exhaust with average ‘‘ambient’’ air.

each other). However, when the two end-members differ in bulk isotopic composition, the mixing line for D47 deviates significantly from linear (Fig. 3B). If one were to fit a straight line through the data (or the correctly calculated mixing curve), one would erroneously conclude that the heated end-member has a D47 value of 0.166 ± 0.083, rather than the actual value of 0.0&. The actual trend can be reasonably well fitted using a second order polynomial. A similar approach was used to estimate the deviation from linearity in the mixing of car exhaust into air, in which the difference in bulk composition between the two end-members is smaller than that in Fig. 3B. The car

where R47, R46, and R45 are the ratio values of 47/44, 46/44, and 45/44 as derived from the measured d47, d46, and d45 for each sample (see Appendix B for an example of the calculation). R47*, R46*, and R45* are the corresponding ratio values calculated assuming random distribution among isotopologues and are derived as follows: R45 ¼ R13 þ 2R17 ; 2

R46 ¼ 2R18 þ 2R13  R17 þ ðR17 Þ ;

ðA2Þ 17 2

R47 ¼ 2R13  R18 þ 2R17  R18 þ R13  ðR Þ ; where R13 and R18 are derived from the measured d13C and d18O, and R17 is derived from R18, assuming mass dependent relation between them. Then " R47 D47 ¼ 2R13  R18 þ 2R17  R18 þ R13  ðR17 Þ2 # R46 R45  18  13 þ 1  1000. 2 17 2R þ 2R13  R17 þ ðR17 Þ R þ 2R ðA3Þ The ÔKeeling plotÕ method is based on mass balance using ca ¼ cb þ cs

and

da c a ¼ d b c b þ ds c s ;

ðA4Þ

where ca is the amount of CO2 (depicted as mixing ratio) in an air sample, cb is the amount of CO2 in an air parcel of background atmosphere, and cs is the amount of CO2 added by

10

H.P. Affek, J.M. Eiler 70 (2006) 1–12

the studied source, in this case by car exhaust. da, db, and ds are the respective d values. The combination results in cb da ¼ ðdb  ds Þ þ ds ðA5Þ ca that is traditionally plotted as da as a function of 1/ca (as in Fig. 1) but can also be plotted as a function of cb/ca, which is the fraction of background air in the measured air (defined as f in Fig. 3). This mass balance approach can be applied to D47 by transforming Eq. (A5) to ratios Ra ¼ f ðRb  Rs Þ þ Rs

ðA6Þ

and substituting Eq. (A6) to each of the ratios comprising Eq. (A3) 47 47 13 13 13 47 ¼ fðf ðR47 a  Rs Þ þ Rs Þ=½2ðf ðRa  Rs Þ þ Rs Þ 18 18 17 17 17  ðf ðR18 a  Rs Þ þ Rs Þ þ 2ðf ðRa  Rs Þ þ Rs Þ 18 18 13 13 13  ðf ðR18 a  Rs Þ þ Rs Þ þ ðf ðRa  Rs Þ þ Rs Þ 2

17 17 46 46 46  ðf ðR17 a  Rs Þ þ Rs Þ   ðf ðRa  Rs Þ þ Rs Þ 18 18 17 17 17 = ½2ðf ðR18 a  Rs Þ þ Rs Þ þ ðf ðRa  Rs Þ þ Rs Þ

2

13 13 17 17 17 þ 2ðf ðR13 a  Rs Þ þ Rs Þðf ðRa  Rs Þ þ Rs Þ 45 45 13 13 13  ðf ðR45 a  Rs Þ þ Rs Þ=½ðf ðRa  Rs Þ þ Rs Þ 17 17 þ 2ðf ðR17 a  Rs Þ þ Rs Þ þ 1g  1000.

ðA7Þ

This cannot be expressed as a simple function of f and must be approximated by expanding it to a polynomial series. The examples shown in Figs. 3B and C indicate that the third and fourth order coefficients are negligible, suggesting that D47 vs. f can be approximated as a second order polynomial fit, in which the intercept equals D47s. Fitting a second order polynomial to the D47 car exhaust ÔKeeling plotÕ yields an end-member value of 0.39& (D47 = 20515[CO2]2 + 232.3[CO2]1 + 0.39, R2 = 0.907). End-members can be estimated from the ÔKeeling plotsÕ also by a regression analysis that takes into account measurement errors in both the isotopic composition (the errors are given in Table 1) and mixing ratios (estimated as 1% of the value). This yielded end-member values of 24.4&, 29.7&, 6.3&, and 0.38& for d13C, d18O, d47, and D47, respectively. For human breath ÔKeeling plotsÕ this regression analysis yielded end-member values of 23.3&, 34.5&, 13.6&, and 0.72& for d13C, d18O, d47, and D47, respectively. In both cases, these values are not significantly different from standard regression analysis (as reported in the body of the text). Appendix B. An example of calculating D47 and d47 To illustrate the methodology for calculation of D47 and d47, we will present here a detailed example of such calculations for one sample (one of the mass spectrometry runs for the 4th ambient air sample in Table 1). From the mass spectrometric measurement we obtain raw values for d45,

d46, and d47. For this sample these values were 5.678, 1.510, and 4.237&, respectively, relative to our working reference gas whose d13CVPDB = 3.698& and d18OVSMOW = 39.272& (as determined by NBS-19 measurements). An 17O correction was performed based on 0:5164 18 18 R17 ¼ R17 , where R17 VSMOW ðR =RVSMOW Þ VSMOW ¼ 18 0:00037995 and RVSMOW ¼ 0:0020052 (Gonfiantini et al., 1993), by varying R13 and R18 (using the Solver function in Excel) to yield a minimal value to the function (R13+2R18  R45)2 + (2R18 + 2R13R17 + (R17)2  R46)2. The resulting d13CVPDB and d18OVSMOW values were 9.801 and 40.856&, respectively. The values of d13C and d18O were then used to calculate 13 R and R18 resulting in 0.0111271 and 0.0020871, respec18 tively (using R13 VPDB ¼ 0:0112372 and RVSMOW ¼ 0:0020052). 17 R was then calculated as above resulting in 0.00038784. The abundance of each one of the isotopes was calculated as: [12C] = (1+R13)1 = 0. 9889954, [13C] = R13/(1 + R13) = 0.0110046, [16O] = (1 + R17 + R18)1 = 0.9975311, 17 17 17 [ O] = R /(1 + R + R18) = 0.00038688, [18O] = R18/ 17 18 (1 + R + R ) = 0.0020820. These abundance values are used to calculate the randomly distributed abundance of the CO2 isotopologues: [44]* = [12][16][16] = 0.9841181, [45]* = [13][16][16] + 2[12][16][17] = 0.011714, [46]* = 2[12] [16][18] + [12][17][17] + 2[13][16][17] = 0.0041166, [47]* = 2[13][16][18] + [13][17][17] + 2[12][17][18] = 4.7304 · 105. The randomly distributed isotopic ratios are then calculated by dividing the respective abundance by [44]*, yielding R45* = 0.0119027, R46* = 0.0041830, and R47* = 5 4.8068 · 10 . The same calculations were done for the refer46 ence gas, yielding R45 ref ¼ 0:011780; Rref ¼ 0:0041101; 5 47 and Rref ¼ 4:7510  10 . Then, the measured isotopic ratios for the samples are obtained from Ri ¼ ðdi=1000þ 1Þ  Ri ref , where i refers to masses 45, 46, and 47, and D47 = [(R47/R47*  1)  (R46/R46*  1)  (R45/R45*  1)] · 1000 = 0.143&. The values are then corrected for the presence of N2O with D14 = 0.00273. The calibration slopes were 62.3, 89.6, and 28.7 for d13C, d18O, and D47, respectively, resulting in d13C = 9.631&, d18O = 41.100&, and D47 = 0.065&. Next, the D47 value obtained was normalized using a standard whose D47 = 0 by definition, which was produced by heating CO2 to 1000 C for 2 h to obtain random distribution of the isotopologues. D47 obtained for this heated gas standard was D47 = 0.824&, and the normalized sample value is therefore D47 = 0.065  (0.824) = 0.889&. Each sample was typically measured three times and the values were averaged. For this sample this resulted in d13C = 9.653&, d18O = 41.078&, and D47 = 0.867& as in Table 1. The scale of d47 was defined here as d47 ¼ 47 47 ðR47 =R47 std  1Þ  1000, where Rstd is the R value for a hypo13 18 thetical CO2 whose d CVPDB = 0, d OVSMOW = 0, and D47 = 0. R47* for the measured heated gas and the hypothetical zero standard were calculated as described above, result5 5 ing in R47 and R47 heated ¼ 4:8248  10 zero ¼ 4:6591  10 . The measured d45, d46, and d47 of the sample were corrected as

Mass 47 CO2 in urban air, car exhaust, and human breath

well for N2O presence using calibration slopes of 59.9, 84.6, and 113.7 for d45, d46, and d47, respectively, resulting in d45 = 5.515&, d46 = 1.741&, and d47 = 3.927&. The calibrated sample d47 was then calculated as ½ðd47= 47 1000 þ 1Þ=ðd47heated =1000 þ 1Þ  ðR47 heated =Rzero Þ  1  1000 ¼ 30:022‰ (d47heated was 1.475&) and the averaged value for the mass spectrometer measurements of this sample was 32.977&. References Andres, R.J., Marland, G., Boden, T., Bischof, S., 2000. Carbon dioxide emissions from fossil fuel consumption and cement manufacture, 1751–1991, and an estimate of their isotopic composition and latitudinal distribution. In: Wigley, T.M.L., Schimel, D.S. (Eds.), The Carbon Cycle. Cambridge University Press, Cambridge, pp. 53–62. Ciais, P., Denning, A.S., Tans, P.P., Berry, J.A., Randall, D.A., Collatz, G.J., Sellers, P.J., White, J.W.C., Trolier, M., Meijer, H.A.J., Francey, R.J., Monfray, P., Heimann, M., 1997. A three-dimensional synthesis study of d18O in atmospheric CO2. 1. Surface fluxes. Journal of Geophysical Research-Atmospheres 102 (D5), 5857–5872. Ciais, P., Tans, P.P., Trolier, M., White, J.W.C., Francey, R.J., 1995a. A large northern-hemisphere terrestrial CO2 sink indicated by the 13 C/12C ratio of atmospheric CO2. Science 269 (5227), 1098–1102. Ciais, P., Tans, P.P., White, J.W.C., Trolier, M., Francey, R.J., Berry, J.A., Randall, D.R., Sellers, P.J., Collatz, J.G., Schimel, D.S., 1995b. Partitioning of ocean and land uptake of CO2 as inferred by d18O measurements from the NOAA Climate Monitoring and Diagnostics Laboratory Global Air Sampling Network. Journal of Geophysical Research-Atmospheres 100 (D3), 5051–5070. Clark-Thorne, S.T., Yapp, C.J., 2003. Stable carbon isotope constraints on mixing and mass balance of CO2 in an urban atmosphere: Dallas metropolitan area, Texas, USA. Applied Geochemistry 18 (1), 75–95. Coplen, T.B., Hopple, J.A., Bohlke, J.K., Peiser, H.S., Rieder, S.E., Krouse, H.R., Rosman, K.J.R., Ding, T., Vocke Jr, R.D., Revesz, K.M., Lamberty, A., Taylor, P., De Bievre, P., 2002. Compilation of minimum and maximum isotope ratios of selected elements in naturally occurring materials and reagents, US Geological Survey Water-Resources Investigations Report 01-4222. pp. 38. Cuntz, M., Ciais, P., Hoffmann, G., Allison, C.E., Francey, R.J., Knorr, W., Tans, P.P., White, J.W.C., Levin, I., 2003. A comprehensive global three-dimensional model of d18O in atmospheric CO2: 2. Mapping the atmospheric signal. Journal of Geophysical Research-Atmospheres 108 (D17), art. no.-4528. De Niro, M.J., Epstein, S., 1978. Influence of diet on distribution of carbon isotopes in animals. Geochimica et Cosmochimica Acta 42 (5), 495–506. Eiler, J.M., Schauble, E., 2004. 18O13C16O in earthÕs atmosphere. Geochimica et Cosmochimica Acta 68 (23), 4767–4777. Epstein, S., Zeiri, L., 1988. Oxygen and carbon isotopic compositions of gases respired by humans. Proceedings of the National Academy of Sciences of the United States of America 85 (6), 1727–1731. Francey, R.J., Tans, P.P., 1987. Latitudinal variation in O-18 of atmospheric CO2. Nature 327 (6122), 495–497. Francey, R.J., Tans, P.P., Allison, C.E., Enting, I.G., White, J.W.C., Trolier, M., 1995. Changes in oceanic and terrestrial carbon uptake since 1982. Nature 373 (6512), 326–330. Friedman, I., OÕNeil, J.R., 1977. Compilation of stable isotope fractionation factors of geological interest. In: Data of Geochemistry, 6th ed. Geological Survey Professional Paper 440-KK. Gonfiantini, R., Stichler, W., Rozanski, K., 1993. Standards and intercomparison materials distributed by the international atomic energy agency for stable isotope measurements. In: Reference and Intercomparison Materials for Stable Isotopes of Light Elements, pp. 13–29. IAEA-TECDOC-825.

11

Ishizawa, M., Nakazawa, T., Higuchi, K., 2002. A multi-box model study of the role of the biospheric metabolism in the recent decline of d18O in atmospheric CO2. Tellus Series B-Chemical and Physical Meteorology 54 (4), 307–324. Kaiser, J., Rockmann, T., Brenninkmeijer, C.A.M., 2003. Assessment of 15 15 16 N N O as a tracer of stratospheric processes. Geophysical Research Letters 30 (2), doi:10.1029/2002GL016253. Keeling, C.D., 1958. The concentration and isotopic abundances of atmospheric carbon dioxide in rural areas. Geochimica et Cosmochimica Acta 13 (4), 322–334. Keeling, C.D., 1961. The concentration and isotopic abundances of carbon dioxide in rural and marine air. Geochimica et Cosmochimica Acta 24 (3-4), 277–298. Kroopnick, P., Craig, H., 1972. Atmospheric oxygen—isotopic composition and solubility fractionation. Science 175 (4017), 54–55. Luz, B., Kolodny, Y., Horowitz, M., 1984. Fractionation of oxygen isotopes between mammalian bone phosphate and environmental drinking water. Geochimica et Cosmochimica Acta 48 (8), 1689– 1693. Miller, J.B., Tans, P.P., 2003. Calculating isotopic fractionation from atmospheric measurements at various scales. Tellus Series B-Chemical and Physical Meteorology 55 (2), 207–214. Miller, J.B., Yakir, D., White, J.W.C., Tans, P.P., 1999. Measurement of 18 O/16O in the soil-atmosphere CO2 flux. Global Biogeochemical Cycles 13 (3), 761–774. Mook, W.G., Jongsma, J., 1987. Measurement of the N2O correction for 13 C/12C ratios of atmospheric CO2 by removal of N2O. Tellus Series B—Chemical and Physical Meteorology 39, 96–99. Mook, W.G., Van der Hoek, S., 1983. The N2O correction in the carbon and oxygen isotopic analysis of atmospheric CO2. Isotope Geoscience 1 (3), 237–242. Na, K., Kim, Y.P., Moon, I., Moon, K.C., 2004. Chemical composition of major VOC emission sources in the Seoul atmosphere. Chemosphere 55 (4), 585–594. Newman, S., Xu, X., Rodrigues, C., Epstein, S., Stolper, E., 1999. Concentrations and isotopic ratios of CO2 in air in the Los Angeles basin as measures of pollution. Eos Trans. AGU, 80, Fall meeting supplement, Abstract A52D-05. Panteleev, N., Peronnet, F., Hillaire-Marcel, C., Lavoie, C., Massicotte, D., 1999. Carbon isotope fractionation between blood and expired CO2 at rest and exercise. Respiration Physiology 116 (1), 77–83. Pataki, D.E., Bowling, D.R., Ehleringer, J.R., 2003a. Seasonal cycle of carbon dioxide and its isotopic composition in an urban atmosphere: anthropogenic and biogenic effects. Journal of Geophysical ResearchAtmospheres 108 (D23). Art. no.-4735. Pataki, D.E., Ehleringer, J.R., Flanagan, L.B., Yakir, D., Bowling, D.R., Still, C.J., Buchmann, N., Kaplan, J.O., Berry, J.A., 2003b. The application and interpretation of Keeling plots in terrestrial carbon cycle research. Global Biogeochemical Cycles 17 (1). Art. no.-1022. Peylin, P., Ciais, P., Denning, A.S., Tans, P.P., Berry, J.A., White, J.W.C., 1999. A 3-dimensional study of d18O in atmospheric CO2: contribution of different land ecosystems. Tellus Series B—Chemical and Physical Meteorology 51 (3), 642–667. Rahn, T., 2005. Factors influencing the stable isotopic content of atmospheric N2O. In: Flanagan, L.B., Ehleringer, J.R., Pataki, D.E. (Eds.), Stable Isotopes and Biosphere–atmosphere Interactions: Processes and Biological Control. Elsevier Academic Press, New York, pp. 268– 287. Reed, B.C., 1992. Linear least-squares fits with errors in both coordinates. 2. Comments on parameter variances. American Journal of Physics 60 (1), 59–162. Takahashi, H.A., Konohira, E., Hiyama, T., Minami, M., Nakamura, T., Yoshida, N., 2002. Diurnal variation of CO2 concentration, D14C and d13C in an urban forest: estimate of the anthropogenic and biogenic CO2 contributions. Tellus Series B-Chemical and Physical Meteorology 54 (2), 97–109.

12

H.P. Affek, J.M. Eiler 70 (2006) 1–12

Tans, P.P., 1981. 13C/12C of industrial CO2. In: Bolin, B. (Ed.), SCOPE 16: Carbon Cycle Modelling. Wiley, New York, pp. 127–129. Wang, Y., Huang, Y.S., 2003. Hydrogen isotopic fractionation of petroleum hydrocarbons during vaporization: implications for assessing artificial and natural remediation of petroleum contamination. Applied Geochemistry 18 (10), 1641–1651. Wang, Z., Schauble, E.A., Eiler, J.M., 2004. Equilibrium thermodynamics of multiply-substituted isotopologues of molecular gases. Geochimica et Cosmochimica Acta 68 (23), 4779–4797. Widory, D., Javoy, M., 2003. The carbon isotope composition of atmospheric CO2 in Paris. Earth and Planetary Science Letters 215 (1–2), 289–298. Yakir, D., Wang, X.F., 1996. Fluxes of CO2 and water between terrestrial vegetation and the atmosphere estimated from isotope measurements. Nature 380 (6574), 515–517.

Yeh, H.W., Epstein, S., 1981. Hydrogen and carbon isotopes of petroleum and related organic matter. Geochimica et Cosmochimica Acta 45 (5), 753–762. Zimnoch, M., Florkowski, T., Necki, J., Neubert, R., 2004. Diurnal variability of d13C and d18O of atmospheric CO2 in the urban atmosphere of Krakow, Poland. Isotopes in Environmental and Health Studies 40 (2), 129–143. Zondervan, A., Meijer, H.A.J., 1996. Isotopic characterization of CO2 sources during regional pollution events using isotopic and radiocarbon analysis. Tellus Series B-Chemical and Physical Meteorology 48 (4), 601– 612. Zyakun, A.M., 2003. What factors can limit the precision of isotope ratio measurements in the isotopomer pool? International Journal of Mass Spectrometry 229 (3), 217–224.